Fig. 5: Comparison of junior radiologists without and with AI assistance.
From: Deep learning approach for screening neonatal cerebral lesions on ultrasound in China
a, b Improvement of metrics (sensitivity, specificity, AUC) with AI assistance on internal test set and external test set. Green color means improvement, and red color means decline. c The diagnostic performance of junior radiologists with AI assistance on specific conditions. d The consistency of junior radiologists without and with AI assistance. Box plots show Cohen’s kappa values between AI and junior radiologists (n = 9 radiologists, biological replicates). Box plots display median (center line), 25th–75th percentiles (box bounds), whiskers extending to the minimum and maximum values within 1.5 × IQR from the quartiles, and individual data points overlaid. Each colored point represents one physician’s diagnostic agreement with AI. The left column represents consistency without AI assistance, while the right column represents consistency with AI assistance. Source data are provided as a Source data file. R radiologist, PVL periventricular leukomalacia.
